In wake of an ever increasing complexity the desire to move towards intelligently controlling operations is amplified in manufacturing. Complex job shops mark the most complex production environments that require a high degree of agility to control. Among these complex manufacturing environments semiconductor manufacturing stands out as it combines all complexities to form a truly complex job shop. Hence, operational excellence is the key to success and relies on intelligent production control. A major concern in controlling such complex job shops, in this case semiconductor wafer fabrication, is the presence of time-constraints that limit the transition time of products between two, mostly successive, processes. Adhering to these product specific time-constraints is of utmost importance as violations result in scrapping the violating product. The state-of-the-art production control of these dispatching decisions that aim at adhering to time-constraints is based on error-prone manual control that is stressful for human operators. Thus, within this thesis a novel, real-time data based approach for intelligently controlling production control for time-constrained complex job shops is presented. Using an up-to-the-minute replica of the real system both uni-, multi-variate time series models and a digital twin are used to obtain violation predictions. As a second step, based on the time-constraint violation expectancy the production control is derived and implemented with a real-world semiconductor manufacturing plant real-time data. The resulting approach is, therefore, validated against the state-of-the-art showing significant improvements as many time-constraint violations could be prevented.